Modeling And Implementation Of Equipment Sudden Large Failure Prediction Based On On‐line Monitoring Data

Semiconductor industry has a strong dependence on its equipment. The equipment with overload operation occurs sudden failure inevitably. It will bring serious damage to the modern production system. There is no obvious sign before the equipment sudden large failure appears. Because of the complexity of equipment, the incompleteness of knowledge and the limitations of detection, it’s hard to forecast or diagnose equipment sudden large failure effectively.To solve this problem, a method of equipment sudden large failure prediction based on on-line monitoring data was proposed. The research was based on a National Natural Science Foundation project called Self-organized Critical State Identification and Risk Research of Machine Sudden Large Failure(Grant No.:51075060). Firstly a fitting model was built for monitoring data of the normal state equipment. Then the residual was used to identify the abnormal situation of equipment and find out the residual critical point which mark the self-organization critical state of equipment. Finally a control chart for residuals was established combined with the equipment failure self-organized critical characteristics. And the early warning of sudden large failure was achieved through the criterion on control chart. A wire bonding machine and an injection molding machine were selected for this research. The bonding temperature and cavity pressure on-line data were used to build ARMA prediction models for the normal state. The residuals of the model verified the SOC characteristics of equipment failure. The mechanism of equipment sudden large failure was explained by applying the SOC theory. Two control lines were set. One was based on the whole residuals through the Shewhart method and the other was based on residuals over the critical point through the weighted variance method. The result of case study showed that this method was valid to forecast the equipment sudden large failure. The SOC control line was better than the Shewhart control line.This thesis explains the equipment sudden large failure phenomenon from a new point and provides a new ideal to analyze equipment failure. The monitoring system will be more useful for the equipment manager. The reliability and maintainability of equipment will be improved. Doing preventive maintenance based on this equipment sudden large failure prediction can ensure the best maintenance chance and avoid the serious accident the sudden large failure brings.